Research map / Image generation / Super-resolution
Dual Aggregation Transformer for Image Super-Resolution
Zheng Chen, Yulun Zhang, Jinjin Gu, Linghe Kong, Xiaokang Yang, Fisher Yu
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| alternate spatial-window and channel-wise self-attention in consecutive blocks (ablation) | DAT · Urban100 x2 | PSNR (dB): 33.20 -> 33.34 | Improved |
| use spatial window SA instead of channel-wise SA only (ablation) | DAT · Urban100 x2 | PSNR (dB): 32.80 -> 33.20 | Improved |
| add parallel depth-wise convolution to self-attention (ablation) | DAT baseline · Urban100 x2 | PSNR (dB): 33.34 -> 33.41 | Improved |
| add adaptive interaction module (AIM) between conv and attention branches (ablation) | DAT baseline · Urban100 x2 | PSNR (dB): 33.41 -> 33.52 | Improved |
| use both interaction directions in AIM instead of one (ablation) | AIM · Urban100 x2 | PSNR (dB): 33.47 -> 33.52 | Improved |
| replace FFN with SGFN (ablation) | DAT · Urban100 x2 | PSNR (dB) / Params: 33.52 -> 33.57, params 16.84M -> 14.66M | Improved |
| add depth-wise convolution in SGFN (ablation) | SGFN · Urban100 x2 | PSNR (dB): 33.44 -> 33.57 | Improved |
| add channel split in SGFN (ablation) | SGFN · Urban100 x2 | PSNR (dB) / Params: 33.53 -> 33.57, params 17.15M -> 14.66M | Improved |
| mix DSTB and DCTB blocks instead of a single block type (ablation) | DAT · Urban100 x2 | PSNR (dB): 33.43 -> 33.57 (vs DSTB only) | Improved |
| use DAT versus CAT-A | DAT · Urban100 x4 | PSNR (dB): 27.89 -> 27.87 | Hurt |
| use DAT-S instead of SwinIR | DAT-S · Urban100 x4 | PSNR (dB) / FLOPs: 27.45 -> 27.68, FLOPs 215.32G -> 203.34G | Improved |
Closest papers by meaning
- Activating More Pixels in Image Super-Resolution Transformer (CVPR 2023 · similarity 0.90)
- Recursive Generalization Transformer for Image Super-Resolution (ICLR 2024 · similarity 0.89)
- Transcending the Limit of Local Window: Advanced Super-Resolution Transformer with Adaptive Token Dictionary (CVPR 2024 · similarity 0.88)
- Efficient Attention-Sharing Information Distillation Transformer for Lightweight Single Image Super-Resolution (AAAI 2025 · similarity 0.87)
- Effective Diffusion Transformer Architecture for Image Super-Resolution (AAAI 2025 · similarity 0.86)
- Progressive Focused Transformer for Single Image Super-Resolution (CVPR 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
